Re: [PATCH net 10/10] selftests: mptcp: explicitly trigger the listener diag code-path

[Matthieu Baerts <matttbe@kernel.org>]

Hi Jakub,

On 01/03/2024 16:10, Matthieu Baerts wrote:
> On 01/03/2024 15:37, Jakub Kicinski wrote:
>> On Fri, 23 Feb 2024 17:14:20 +0100 Matthieu Baerts (NGI0) wrote:
>>> From: Paolo Abeni <pabeni@redhat.com>
>>>
>>> The mptcp diag interface already experienced a few locking bugs
>>> that lockdep and appropriate coverage have detected in advance.
>>>
>>> Let's add a test-case triggering the relevant code path, to prevent
>>> similar issues in the future.
>>>
>>> Be careful to cope with very slow environments.
>>>
>>> Note that we don't need an explicit timeout on the mptcp_connect
>>> subprocess to cope with eventual bug/hang-up as the final cleanup
>>> terminating the child processes will take care of that.
>>
>> Hi!
>>
>> There's a failure in CI under debug after merging net and net-next 
>> in diag.sh. Maybe because of the patch which lowered timeout?
>> https://lore.kernel.org/all/20240223-upstream-net-next-20240223-misc-improvements-v1-8-b6c8a10396bd@kernel.org/
> 
> Thank you for this message!
> 
> I didn't have this error on my side, even without '-d SLUB_DEBUG_ON' we
> do on top of the debug kconfig, but I see I can reproduce it on slower
> environments. Indeed, it looks like it can be caused by that
> modification. I will send a fix ASAP!

The following patch fixes the issue on my side (when using 'renice 20'
and stress-ng in parallel :) ):

https://lore.kernel.org/netdev/20240301-upstream-net-20240301-selftests-mptcp-diag-exit-timeout-v1-2-07cb2fa15c06@kernel.org/T/

I queued it for -net, but the issue should only be visible in net-next
due to the reduction of the timeout, as you suggested. This patch can
also be applied in net-next if preferred, it will not cause any
conflicts between net and net-next.

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.